Seismic impedance inversion is an effective method for oil and gas reservoir prediction and has been widely used in industry. Although the technique is very mature, it still faces the problems such as strong ill posedness, poor lateral continuity, and low resolution. The structure-oriented regularization proposed by predecessors can effectively reduce ill posedness in the least-squares inversion process and enhance the lateral continuity of the inversion results, but it cannot fully preserve the geological boundaries and may even cause damage to discontinuous signals such as faults. In order to enhance the ability of inversion algorithms to depict geological structure details and highlight the "block" characteristics of the impedance model, we introduce edge-preserving filter during the inversion iteration process, and use a weight parameter to control the sharpness of model update. For each sampling point in the impedance model, edge-preserving filter searches for the most uniform window around it and assigns the average value of all samples within that window to that point. This processing can preserve sharp interfaces from blurring and significantly improve the resolution of inversion results. The results of model and field data testing show that the proposed method can significantly improve the ability to characterize the structure boundaries and has excellent performance in enhancing the discontinuity features of impedance inversion results.